A method and lifting device for visual double-box detection of containers

Abstract

The present invention relates to the field of container systems, and specifically to a method and a sling for visual double-box detection of containers, comprising the following steps: (1) using a camera device to shoot a video of the middle of a container, and transmitting the video to a sling system; (2) converting the video into a picture in the sling system, and marking the lock hole of the container in the picture; (3) using a recognition model to identify the lock hole of the container; (4) if the lock hole is not identified, continuing the lifting operation, and if the lock hole is identified, stopping the lifting operation. The method of marking the lock hole and using AI for identification can shorten the judgment time, give full play to the user's ability, and maximize the operating efficiency. At the same time, it does not require complex circuits, and the accuracy of manual identification is higher.

Description

Technical Field

[0001] The present invention relates to the field of container systems, and in particular to a container visual double-box detection method and a lifting device. Background Art

[0002] Special rigging such as hooks, wire ropes, and chains are commonly used in container loading and unloading. Lifting suction cups, clamps, and forks can be used as special container loading and unloading spreaders on cranes for a long time, or they can be hung on hooks as replaceable auxiliary container loading and unloading spreaders for temporary use. They are often used in warehouses and yards with multiple cargo types to improve operating efficiency. The container loading and unloading spreader is connected to the top corner accessories of the container through the twist locks at the four corners of its end crossbeam, and the driver controls the opening and closing of the twist locks to carry out container loading and unloading operations.

[0003] The prior art has the following problems: double-box detection is performed by manual reminder or photoelectric switch to detect the box seam between the double boxes. Manual reminder will reduce the operating efficiency and is prone to misjudgment, which may lead to danger; photoelectric switch is used to detect the box seam between the double boxes, which requires a large number of photoelectric switches and has a complex internal circuit. The photoelectric switch is easily interfered with and misjudgment may occur. Summary of the invention

[0004] Therefore, it is necessary to provide a container visual double-box detection method and a sling, which solves the problems that the reminder reduces the operating efficiency and the double-box detection circuit of the photoelectric switch is complex and susceptible to interference.

[0005] To achieve the above object, the present invention provides a method for visual double-box detection in a container, comprising the following steps:

[0006] S10 uses a camera to shoot video of the middle of the container and transmits the video to the spreader’s system;

[0007] S20 converts the video into an image in the spreader’s system and marks the keyholes of the container in the image;

[0008] S30 trains a recognition model and uses the recognition model to identify the keyhole of the container;

[0009] If the keyhole is not identified in S40, the hoisting operation is continued; if the keyhole is identified, the hoisting operation is stopped.

[0010] Furthermore, in step S30, the training process of the recognition model is:

[0011] S31 collects operation videos, converts them into pictures to form a data set, uses the Labelme tool to mark the keyhole area in the data set with a rectangle, and then exports the label file as the label file for simulation training.

[0012] The YOLOV5 deep learning model adopted by S32 obtains the rectangular information of the label file, and clusters 9 Anchors through the K-means clustering method. The Anchor is the length and width of the rectangle. The model inputs the picture, and the model outputs the predicted position of the keyhole. The data output by the model and the data of the label file are used as the loss function. Finally, the reverse gradient optimization method is used to make the model prediction result infinitely close to the data of the label file.

[0013] S33 continuously adjusts the training hyperparameters and supplements the data to finally obtain the recognition model.

[0014] Furthermore, in step S31, the labeled data set includes a training set and a validation set. The purpose of the training set is to train the model, and the purpose of the validation set is to verify the accuracy of the model and prevent overfitting of the model training.

[0015] Furthermore, the method further comprises step S50, wherein step S50 is:

[0016] Angle sensors are used at both ends of the spreader to detect the angle between the container and the horizontal plane. If the angle between the container and the horizontal plane exceeds the angle threshold, the lifting operation is stopped. If the angle between the container and the horizontal plane is within the angle threshold, the lifting operation is continued.

[0017] Furthermore, the angle threshold is 85-95°.

[0018] The present invention also discloses a sling applying the above-mentioned container visual double-box detection method, comprising a main frame, two telescopic beams, two connecting beams and two angle detection devices, the two telescopic beams are relatively arranged on both sides of the main frame, the two telescopic beams are movably connected to the main frame, the two connecting beams are respectively fixed on one telescopic beam, both ends of the connecting beams are fixed with a sling lock, the two angle detection devices are respectively fixed on one connecting beam, and a camera device is fixed in the middle of the main frame, and the camera device is vertically facing the bottom of the main frame.

[0019] Furthermore, the camera device is externally connected to a control panel, and the control panel is electrically connected to the angle detection device. The control panel transmits the information captured by the camera device and the angle information of the angle detection device to the user, thereby helping the user to judge the number of containers and avoid hoisting two containers mistakenly as one container.

[0020] Furthermore, the angle detection device includes a detection plate and an angle sensor, the detection plate is rotatably connected to the connecting beam, and the detection plate is vertically arranged, the angle sensor is fixed on the side of the detection plate away from the main frame, and the angle sensor is electrically connected to the control panel. The angle detection device is a second double-box detection. When the information captured by the camera device cannot determine whether it is a double box or it is considered that the key information is missed due to an error, resulting in the lifting of the double box, when the lifting lock and the container are engaged with each other, the detection plate is located on the side of the container, and after one end of the container is lifted, the angle of the detection plate changes, and the angle sensor transmits the angle information to the control panel, which serves as a warning to the user, so that the user can find the error in time and stop lifting the double box.

[0021] Furthermore, the main frame includes two fixed plates and two sliding plates, the two fixed plates are arranged opposite to each other, the two sliding plates are fixed between the two fixed plates, the camera device is arranged in the space enclosed by the fixed plates and the sliding plates, the two ends of the camera device are fixedly connected to the sides opposite to the fixed plates, the fixed plates are provided with through holes, the telescopic beam passes through the through holes and is slidably connected to the sliding plates. The main frame is enclosed by a plate-like structure, which reduces the weight of the main frame to the maximum extent without affecting the firmness of the main frame, reduces the weight of the overall sling, and reduces the energy consumption of the sling's own weight.

[0022] Furthermore, the telescopic beam includes a bottom plate and two side plates, the two side plates are fixed on two opposite sides of the bottom plate, the opposite sides of the sliding plate are provided with slide grooves, and the side plates of the telescopic beam are fixed with slide bars matching the slide grooves, and the slide bars are slidably connected with the slide grooves. The telescopic beam is enclosed by a plate-like structure, which can reduce the weight of the telescopic beam to the maximum extent without affecting the firmness of the telescopic beam, reduce the weight of the overall sling, and reduce the energy consumption of the sling's own weight. The telescopic beam is slidably connected to the main frame, which can flexibly control the unfolded length of the telescopic beam to meet the needs of containers of different sizes and increase the practicality of the sling.

[0023] Furthermore, the connecting beam is fixed on the lower surface of the base plate away from one end of the main frame, and the hanging lock includes a lock column and a lock core fixedly connected to each other. The lower surface of the connecting beam is recessed upward to form an installation groove, the lock column is embedded in the installation groove, and the lock core extends downward to the outside of the installation groove.

[0024] Furthermore, the connecting beam is fixed on the lower surface of the base plate away from one end of the main frame, and the hanging lock includes a lock column and a lock core fixedly connected to each other. The lower surface of the connecting beam is recessed upward to form an installation groove, the lock column is embedded in the installation groove, and the lock core extends downward to the outside of the installation groove.

[0025] Furthermore, a lifting hole is provided on the fixing plate.

[0026] Furthermore, a lighting lamp is fixed on the outer side of the sliding plate, the lighting lamp corresponds to the position of the camera device, and the lighting lamp is vertically downward. The lighting lamp ensures that the camera device can clearly capture the situation of the container in a dark environment.

[0027] Different from the prior art, the above technical solution has the following beneficial effects:

[0028] 1. Marking the keyhole and using AI for identification can shorten the judgment time, give full play to the user's ability, and maximize operational efficiency. At the same time, it does not require complex circuits, and manual identification has a higher accuracy rate.

[0029] 2. The camera device is for the first double-box detection. A camera device is installed in the middle of the main frame. The camera device is used to shoot the upper surface of the container. When two containers are arranged in a row, there will be four lock holes on the upper surface of the two containers. When the camera device shoots four lock holes, it can remind the user to avoid lifting the two containers at the same time; the angle detection device is for the second double-box detection. When the information shot by the camera device cannot determine whether it is a double container or it is believed that the key information is missed due to an error, resulting in the lifting of the double containers, when the lifting lock and the container are engaged with each other, one end of the container is lifted, the angle of the container changes, and the angle sensor transmits the angle information to the user to alert the user so that the user can find the error in time and stop lifting the double containers. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flow chart of the double-box detection method described in Specific Implementation 1;

[0031] Figure 2 This is a schematic diagram of the structure of the sling described in Specific Implementation 2;

[0032] Figure 3 It is an enlarged schematic diagram of the part A of the sling described in the specific implementation mode 2;

[0033] Figure 4 This is a physical diagram of the lifting lock structure of the lifting device described in specific implementation mode 2.

[0034] Description of reference numerals:

[0035] 1. Main frame; 101. Fixed plate; 1011. Lifting hole; 102. Sliding plate; 103. Camera device; 104. Lighting lamp; 105. Fixed column;

[0036] 2. telescopic beam; 201. side plate; 202. bottom plate; 2021. connecting plate; 203. slide bar;

[0037] 3. Connecting beam; 301. Angle detection device; 3011. Angle sensor; 3012. Detection plate; 3013. Rotating shaft; 302. Hanging lock; 3021. Lock core; 3022. Lock column. DETAILED DESCRIPTION

[0038] In order to explain the technical content, structural features, achieved objectives and effects of the technical solution in detail, the following is a detailed description in conjunction with specific embodiments and accompanying drawings.

[0039] Embodiment 1:

[0040] See also Figure 1 This embodiment discloses a method for visual double-box detection in a container, comprising the following steps:

[0041] S10 uses a camera to shoot video of the middle of the container and transmits the video to the spreader’s system;

[0042] S20 converts the video into an image in the spreader’s system and marks the keyholes of the container in the image;

[0043] S30 trains a recognition model and uses the recognition model to identify the keyhole of the container;

[0044] If the keyhole is not identified in S40, the hoisting operation continues; if the keyhole is identified, the hoisting operation stops;

[0045] S50 uses angle sensors 3011 at both ends of the lifting device to detect the angle between the container and the horizontal plane. If it is detected that the angle between the container and the horizontal plane exceeds the angle threshold, the lifting operation is stopped. If it is detected that the angle between the container and the horizontal plane is within the angle threshold, the lifting operation is continued.

[0046] In step S30, the training process of the recognition model is:

[0047] S31 collects operation videos, converts them into pictures to form a data set, uses the Labelme tool to mark the keyhole area in the data set with a rectangle, and then exports the label file as the label file for simulation training.

[0048] The YOLOV5 deep learning model adopted by S32 obtains the rectangular information of the label file, and clusters 9 Anchors through the K-means clustering method. The Anchor is the length and width of the rectangle. The model inputs the picture, and the model outputs the predicted position of the keyhole. The data output by the model and the data of the label file are used as the loss function. Finally, the reverse gradient optimization method is used to make the model prediction result infinitely close to the data of the label file.

[0049] S33 continuously adjusts the training hyperparameters and supplements the data to finally obtain the recognition model.

[0050] In step S31, the labeled data set includes a training set and a validation set. The purpose of the training set is to train the model, and the purpose of the validation set is to verify the accuracy of the model and prevent overfitting of the model training. The angle threshold is 85-95°.

[0051] Embodiment 2:

[0052] See also Figure 2-4 This embodiment discloses a device for applying the above-mentioned container visual double-box detection method, including a main frame 1, two telescopic beams 2, two connecting beams 3, four lifting locks 302, a camera device 103 and two angle detection devices 301; the main frame 1 includes two fixed plates 101 and two sliding plates 102, the two fixed plates 101 are arranged opposite to each other, the two sliding plates 102 are fixed between the two fixed plates 101, the camera device 103 is arranged in the space enclosed by the fixed plates 101 and the sliding plates 102, the camera device 103 vertically illuminates the bottom of the main frame 1, the camera device 103 is externally connected to a control panel, and the control panel transmits the information captured by the camera device 103 to the user, so as to help the user judge the number of containers and avoid two containers being mistaken for one container for lifting. Fixed columns 105 are provided on the opposite sides of the fixed plates 101, and both ends of the camera device 103 are fixedly connected to the fixed columns 105. An illuminating lamp 104 is fixed on the outer surface of the sliding plate 102 . The illuminating lamp 104 corresponds to the position of the camera device 103 . The illuminating lamp 104 illuminates vertically downward. A hoisting hole 1011 is also provided on the fixing plate 101 .

[0053] The telescopic beam 2 includes two side plates 201 and a bottom plate 202, and the two side plates 201 are fixed on the opposite sides of the bottom plate 202. The two telescopic beams 2 are respectively arranged on the sides of the fixed plate 101, and the fixed plate 101 is provided with a through hole, and the opposite sides of the sliding plate 102 are provided with a slide groove, and the side plate 201 is provided with a slide bar 203 matching the slide groove, and the slide bar 203 is slidably connected with the slide groove. The main frame 1 and the telescopic beam 2 are enclosed by a plate-like structure, which can reduce the weight of the main frame 1 to the maximum extent without affecting the firmness of the main frame 1, reduce the weight of the overall sling, and reduce the energy consumption of the sling's own weight. The telescopic beam 2 is slidably connected to the main frame 1, and can flexibly control the unfolded length of the telescopic beam 2 to meet the needs of containers of different sizes and increase the practicality of the sling.

[0054] The two sides of the end of the bottom plate 202 away from the main frame 1 extend outward to form a connecting plate 2021. The connecting plate 2021 is provided with a fastener. In this embodiment, the fastener is a bolt device, and the connecting beam 3 is fixedly connected to the bottom plate 202 through the bolt device. Four hanging locks 302 are respectively fixed on the two ends of the connecting beam 3. The hanging lock 302 includes a lock column 3022 and a lock core 3021 that are fixedly connected to each other. The lower surface of the connecting beam 3 is recessed upward to form an installation groove. The lock column 3022 is embedded in the installation groove, and the lock core 3021 extends downward to the outside of the installation groove.

[0055] The angle detection device 301 includes a detection plate 3012 and an angle sensor 3011. The detection plate 3012 is rotatably connected to the connecting beam 3 through a rotating shaft 3013, and the detection plate 3012 is vertically arranged. The angle sensor 3011 is fixed on the side of the detection plate 3012 away from the main frame 1, and the angle sensor 3011 is electrically connected to the control panel. The angle detection device 301 is a second double-box detection. When the information captured by the camera device 103 cannot determine whether it is a double box or it is considered that the key information is missed due to an error, resulting in the lifting of the double box, when the lifting lock 302 and the container are engaged with each other, the detection plate 3012 is located on the side of the container. After one end of the container is lifted, the angle of the detection plate 3012 changes, and the angle sensor 3011 transmits the angle information to the control panel, which serves as a warning to the user, so that the user can find the error in time and stop lifting the double box.

[0056] When in use, the sling is hoisted through the hoisting hole 1011. The camera device 103 is the first double-box detection. The camera device 103 in the middle of the main frame 1 takes a picture of the container to be hoisted and transmits it to the control panel. The user identifies whether it is a double box through the control panel, thereby avoiding the danger of mistaking the double box for a single box. The angle detection device 301 is the second double-box detection. When the double box is not identified due to human reasons, when the double box is hoisted, the lifting lock 302 is engaged with the lock hole of the container, and the detection plate 3012 of the angle detection device 301 fits with the side of the container. Since the double boxes are arranged in a row, the hook hooks up one side of the container, causing the container to tilt, and then the vertical state of the detection plate 3012 is destroyed. The angle information is transmitted to the control panel through the angle sensor 3011 to alarm the user, avoiding the loss caused by the user lifting the double box without knowing it.

[0057] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "include..." or "comprise..." do not exclude the existence of other elements in the process, method, article or terminal device including the elements. In addition, in this article, "greater than", "less than", "exceed" and the like are understood to exclude the number itself; "above", "below", "within" and the like are understood to include the number itself.

[0058] Although the above embodiments have been described, once those skilled in the art know the basic creative concepts, they can make additional changes and modifications to these embodiments. Therefore, the above description is only an embodiment of the present invention and does not limit the patent protection scope of the present invention. Any equivalent structure or equivalent process transformation made by using the contents of the specification and drawings of the present invention, or directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A double-box detection method, characterized in that: The following steps are involved: S10 uses a camera to shoot video of the middle of the container and transmits the video to the spreader’s system; S20 converts the video into an image in the spreader’s system and marks the keyholes of the container in the image; S30 trains a recognition model and uses the recognition model to identify the keyhole of the container; If the keyhole is not identified in S40, the hoisting operation continues; if the keyhole is identified, the hoisting operation stops; S50 uses angle sensors at both ends of the spreader to detect the angle between the container and the horizontal plane. If it is detected that the angle between the container and the horizontal plane exceeds the angle threshold, the lifting operation is stopped. If it is detected that the angle between the container and the horizontal plane is within the angle threshold, the lifting operation is continued. A lifting device applied to the double-box detection method comprises a main frame, two telescopic beams, two connecting beams and two angle detection devices, wherein the two telescopic beams are relatively arranged on both sides of the main frame, the two telescopic beams are movably connected to the main frame, the two connecting beams are respectively fixed to one telescopic beam, both ends of the connecting beams are fixed with a lifting lock, the two angle detection devices are respectively fixed to one connecting beam, and a camera device is fixed in the middle of the main frame, and the camera device is vertically facing the bottom of the main frame; The camera device is externally connected to a control panel, and the control panel is electrically connected to the angle detection device; The angle detection device includes a detection plate and an angle sensor, wherein the detection plate is rotatably connected to the connecting beam and is vertically arranged, and the angle sensor is fixed on a side of the detection plate away from the main frame, and the angle sensor is electrically connected to the control panel; The main frame includes two fixed plates and two sliding plates, the two fixed plates are arranged opposite to each other, the two sliding plates are fixed between the two fixed plates, the camera device is arranged in a space enclosed by the fixed plates and the sliding plates, the two ends of the camera device are fixedly connected to the side surfaces opposite to the fixed plates, the fixed plates are provided with through holes, the telescopic beam passes through the through holes and is slidably connected to the sliding plates; The lifting device is lifted up through the lifting hole. The camera device is used for the first double-box detection. The camera device in the middle of the main frame takes a picture of the container to be lifted and transmits it to the control panel. The user identifies whether it is a double box through the control panel, thereby avoiding the danger of mistaking the double box for a single box when lifting it. The angle detection device is used for the second double-box detection. When the double boxes are not identified as being arranged in a row due to human reasons, when the double boxes are lifted, the lifting lock engages with the lock hole of the container, and the detection plate of the angle detection device fits with the side of the container. Since the double boxes are arranged in a row, the hook hooks up one side of the container, causing the container to tilt, and then the vertical state of the detection plate is destroyed. The angle information is transmitted to the control panel through the angle sensor to alarm the user, thereby avoiding the loss caused by the user lifting the double boxes without knowing it.

2. The double-box detection method according to claim 1, characterized in that: In step S30, the training process of the recognition model is as follows: S31 collects operation videos, converts them into pictures to form a data set, uses the Label me tool to mark the keyhole area in the data set with a rectangle, and then exports the label file as the label file for simulation training; The YOLOV5 deep learning model adopted by S32 obtains the rectangular information of the label file, and clusters 9 Anchors through the K-means clustering method. The Anchor is the length and width of the rectangle. The model inputs the picture, and the model outputs the predicted position of the keyhole. The data output by the model and the data of the label file are used as the loss function. Finally, the reverse gradient optimization method is used to make the model prediction result infinitely close to the data of the label file. S33 continuously adjusts the training hyperparameters and supplements the data to finally obtain the recognition model.

3. The double-box detection method according to claim 2, characterized in that: In step S31, the labeled data set includes a training set and a validation set.

4. The double-box detection method according to claim 1, characterized in that: The angle threshold is 85-95°.

5. The double-box detection method according to claim 1, characterized in that: The telescopic beam comprises a bottom plate and two side plates, the two side plates are fixed on two opposite sides of the bottom plate, sliding grooves are arranged on opposite sides of the sliding plate, sliding bars matching the sliding grooves are fixed on the side plates of the telescopic beam, and the sliding bars are slidably connected with the sliding grooves.

6. The double-box detection method according to claim 5, characterized in that: The connecting beam is fixed on the lower surface of the base plate away from the main frame. The hanging lock includes a lock column and a lock core that are fixedly connected to each other. The lower surface of the connecting beam is recessed upward to form a mounting groove. The lock column is embedded in the mounting groove, and the lock core extends downward to the outside of the mounting groove.

7. The double-box detection method according to claim 6, characterized in that: Both sides of the bottom plate at one end away from the main frame extend outward to form a connecting plate, and fasteners are arranged on the connecting plate, and the connecting beam is fixedly connected to the bottom plate through the fasteners.

Images